JP2015527658A - Method, system, and computer program product for asynchronous message sequencing in a distributed parallel environment - Google Patents

Abstract

Distributed asynchronous message sequencing by a computer using at least one data processor in a distributed parallel system having a plurality of inbound handlers forming an inbound handler layer and a plurality of outbound handlers forming an outbound handler layer A method in which any of the plurality of inbound handlers receives an incoming message that includes a sequence correlation value that identifies a sequence that includes the incoming message; the sequence status of the sequence in a shared sequence storage; Checking; and determining whether the incoming message is the next message to process in order to maintain the order of the messages in the sequence. And when the sequence status indicates that none of the outbound handlers of the outbound handler layer is currently processing the message of the sequence, and it is determined that the incoming message is a message to be processed next to the sequence, The incoming message is forwarded to queue storage and then retrieved by the outbound handler layer's available outbound handler for processing, and the sequence status is at least one of the outbound handler layer's outbound handlers currently said sequence The queue storage already contains messages to be processed in the sequence, or the incoming message is in the sequence Then when it is determined not to be the message to be processed, and stores the incoming message in the memory of the shared overflow storage, a method of holding for further processing. [Selection] Figure 8B

Description

  The present invention relates generally to data and information processing in communication systems, and more specifically to a method, apparatus, and system for processing a sequence of asynchronous messages in a distributed parallel processing environment.

  When a distributed software architecture is used in service calls or event processing, message transmission is performed synchronously or asynchronously. Messages are delivered and multicast to individual full recipients, and each multicast message is processed individually.

  FIG. 1 shows the synchronous transmission of a message or service call between two systems, one of which is a paging system 110 and the other is a remote system 120, which controls the order of message processing. In this case, the call system 110 waits for the result of the remote processing, so that the caller becomes the master for the order in which messages are actually processed at the server system or remote system.

  Transmission 111 of the first message A from the calling system 110 is processed at the remote system 120 and then returned to the calling system 110 as a processed message A121. Once the processed message A is received, the call system 110 can transmit the second message B 113 to the remote system 120. The second message B is then processed by the remote system 120 and the processed message B 123 is returned to the calling system 110.

  The example shown in this flow chart shows the sequence of synchronous calls or message processing between the calling system 110 and the remote system 120, and the processing 112 of the first message A of the server system or remote system 120 is the second message B. It can be seen that this process is performed before the process 114.

  FIG. 2 shows the asynchronous transmission of a message or service call between the call system 210 and the server system or remote system 220, which sends the service call or message to the server system or remote system 220, and these The system 220 processes messages according to its own scheduling. The client system or call system 210 relaxes control of message processing timing.

  Transmission 211 of the first message A from the calling system 210 is processed by the remote system 220. Meanwhile, paging system 210 has begun transmitting 213 second message B to remote system 220. The second message B is then processed by the remote system 220, but it cannot be determined to return the processed message B to the calling system 210 until message A is processed.

  The example shown in this flowchart shows the order of asynchronous calls or message processing between the call system 210 and the remote system 220, and the processing 212 of the first message A of the server system or remote system 220 is changed to the second message B. It is shown that the process 214 is performed almost simultaneously. Although it is possible to process the second message B before the first message A, it may have a significant impact on the suitability of the sequence containing messages A and B.

  FIG. 3 is a flowchart illustrating parallel processing of service calls or messages in a distributed system. In distributed systems, service calls or messages are processed in parallel and / or multi-threaded by instantiation to meet fault tolerance and scalability requirements. In this figure, instances 1, 2, 3,..., And n corresponding to processing systems 310-1, 310-2,..., And 310-n are represented by four messages 1, 2, 3, and 4 is processed by the inbound sequence.

  Parallel processing does not guarantee the order in which consecutive service calls and messages are processed and completed. However, service calls and message processing may strongly require execution of sequences between related events and messages.

  In the example shown in FIG. 3, message 1 is processed by instance 2, message 2 is processed by instance 3, message 3 is processed by instance 1, and message 4 is processed by instance 3 as message 2. Since message processing is not coordinated, message 2 is processed first, then message 1, message 4, and finally message 3. This is an inconsistent processing order.

  In this figure, the sequence refers to the order for service call and message transmission and / or processing by the distributed system. This order is usually driven by business processing or industry standards. Failure to follow this order results in inadequate processing, which in the worst case results in irreversible corruption of the stored functional data and corruption of the called database.

  FIG. 4 is a flow chart illustrating the parallel processing of asynchronous calls or messages in a distributed processing system and the risk of message processing not being performed in order and the risk of data corruption.

  In a synchronous environment, sequencing is ensured by an emitter system or paging system that sequentially executes message transmissions to remote systems and effectively controls the sequence flow between related messages.

  Since it is not possible to determine the end of message processing at the remote system 420, this sequencing is not possible when the emitter system or paging system 410 needs to handle asynchronous distributed processing. FIG. 4 illustrates this risk of transmission 413 of the second message B following transmission 411 of the first message A from the calling system 410 to the remote system 420. The process 414 of the second message B starts before the process 412 of the process 414 of the first message A. Therefore, the message processing is reversed, resulting in insufficient processing. In the worst case, the stored functional data is irreversibly damaged or the database is damaged.

  Accordingly, an object of the present invention is to alleviate the above-mentioned problems and to avoid irreversible damage to stored functional data and databases.

One embodiment of the present invention performs distributed asynchronous message sequencing on a computer in a distributed parallel system having a plurality of inbound handlers forming an inbound handler layer and a plurality of outbound handlers forming an outbound handler layer. A method is provided that is performed using at least one data processor:
Receiving an incoming message with a sequence correlation value identifying a sequence including the incoming message in any of the plurality of inbound handlers; and checking the sequence status of the sequence in the shared sequence storage; Determining whether the incoming message is the next message to be processed in order to maintain the order of the messages in the sequence;
If the sequence status indicates that none of the outbound handler layer's outbound handlers is currently processing messages in the sequence, and it is determined that the incoming message is a message to be processed next to the sequence, Forward the incoming message to shared queue storage, then get the message by an outbound handler available for processing,
If the sequence status indicates that at least one of the outbound handler layer's outbound handlers is currently processing the message of the sequence, or if the shared queue storage already contains the message to be processed of the sequence Alternatively, if it is determined that the incoming message is not a message to be processed next in the sequence, the incoming message is stored in the shared overflow storage memory and retained for further processing.

  Thus, a distributed parallel system includes: an inbound handler arranged in parallel to each other in an inbound handler layer that receives messages related to many sequences; and a shared sequence storage, a queue storage, and a shared overflow storage from an inbound handler It can be considered a router that includes a storage layer configured to receive messages and store them in memory. The sequence storage may include overflow storage. A distributed parallel system is configured so that it is arranged in parallel with each other in the outbound handler layer and retrieves messages from shared queue storage for processing, ensuring accurate message sequencing within each sequence A further outbound handler. The outbound handler is configured to receive a message for processing and possibly delivery to the correct recipient.

  Overflow storage and sequence storage are shared by inbound handlers and outbound handlers. They are shared and used by parallel inbound handlers and parallel outbound handlers.

  Thus, the present invention provides a solution for maintaining the order of messages for the same sequence while performing parallel processing of various sequences in a distributed environment. Further, separating the inbound handler from the outbound handler makes it possible to separate the emitter throughput from the receiver throughput. In addition, the number of inbound and outbound handlers can be advanced and individually expanded. The present invention also avoids similarities between sequences and inbound or outbound handlers so that inbound or outbound handlers can process any sequence of messages. Therefore, the present invention provides strong fault tolerance because some handlers and outbound handlers do not affect message processing.

  The method of the present invention may further include any of the additional functions and steps shown below.

In one embodiment, to maintain the order of messages in the sequence, the determination of whether an incoming message is the next message to process is:
Determining a message rank indicating the order of incoming messages in the sequence; andcompare the message rank with a sequence rank defining a rank of the next message to be processed in the sequence;
If the message rank is the same as the sequence rank, the message is determined to be the next message to be processed in order to maintain the order of the messages in the sequence;
If the message rank is greater than the sequence rank, the message is not considered the next message to be processed in order to maintain the order of the messages in the sequence.

  Messages that are included in the sequence and are to be processed in a particular order are given a message rank number, which is referred to as a description of the current message rank. Message rank defines the order of messages in a sequence. A message rank may be given to a message by a message originator or a third party. The message rank may be assigned by the system depending on the order of arrival of the messages in the sequence.

  As described herein, the sequence rank defines which message of a given sequence should be processed next, that is, the message rank that the message to be processed next should have.

  The sequence rank is preferably indicated in the sequence storage.

Typically, processing a message by an outbound handler means that the outbound handler sends or delivers the message to the recipient.

Advantageously, once processing of a given sequence of messages in the outbound handler is complete, the sequence rank of the given sequence is incremented.

  When incrementing the sequence rank of a sequence, the method preferably checks whether the overflow storage contains a message whose message rank is incremented and is equal to the sequence rank, and then forwards this message to the queue storage. .

  According to an advantageous embodiment, the step of determining the message rank when the received incoming message is not given an index indicating the message rank in the sequence comprises the step of determining the message rank indicating the rank of the incoming message in the sequence: Assigning to incoming messages and storing the assigned message rank in sequence storage.

The assigned message rank preferably corresponds to the rank of the last message received at any one place in the inbound handler, incremented by adding 1 to the sequence. Therefore, if the incoming message is the first message in the sequence, the message rank is 1. If the message rank of the previous message received by a certain inbound handler is N, the message rank assigned to the newly received message is N + 1.

  In another advantageous embodiment, incoming messages received by the inbound handler are given an index that indicates the message rank in the sequence.

If the message rank is greater than the sequence rank, it is preferable to set the status of the sequence to “pending”. Thus, “pending” means that the overflow storage area contains at least one message of a given sequence, but the message rank of this (these) messages is not equal to the sequence rank.

If no outbound handler is currently processing the sequence message and the sequence message is not in the overflow storage area, typically the sequence status is set to “waiting”. If at least one outbound handler is currently processing a message of the sequence, typically the sequence status is set to “processing”.

  Advantageously, if the queue storage does not contain a message in the sequence of incoming messages, and if the message rank of the incoming message is greater than the sequence rank indicated in the sequence storage, the incoming message is incremented by the sequence rank and the incoming message It is stored in overflow storage until it is equal to the message rank of the message.

Thus, if a message is given a message rank from the message originator or a third party and the message rank is greater than the sequence rank, the message is stored in overflow storage. If other lower message rank messages are processed, the sequence rank is incremented until the message rank reaches the message rank of the previously stored message. This message is then released from sequence storage, more specifically overflow storage, and once queue storage and inbound handlers have not stored and cannot process this sequence of messages, they are sent to queue storage. Sometimes.

The same applies to messages that are not assigned a message rank but are assigned a message rank by the system according to the order of arrival.

  Advantageously, if the message is properly processed by the outbound handler, the message is deleted from the queue storage.

  Advantageously, the outbound handler operates asynchronously, allowing the message to be sent and allowing further processing after sending the message until an acknowledgment is received from the recipient of the message.

  According to an advantageous embodiment, the outbound handler includes a delivery process for sending a message to the recipient and an acknowledgment process for receiving an acknowledgment from the recipient. Since the delivery process and the acknowledgment process operate independently, the delivery process can be used immediately after sending the message.

  Advantageously, until an incoming message is received and proceeds to the checking step, this method ensures that all inbound handlers will receive the rest of the sequence until the incoming message is stored in sequence storage or sent to queue storage. Performing an inbound lock step that is stopped from receiving the message.

Advantageously, incoming messages can be accepted by the inbound handler while other messages in the same sequence are sent and processed by the outbound handler. Situations where an incoming message needs to wait for release from lock are limited to:
If another incoming message is stored in the storage layer or received by the inbound handler; and if the recipient's response to the sequence of messages is received and processed by the outbound handler. If the outbound handler receives a response (acknowledgment) from the recipient, the sequence, the corresponding rank, if any, the time to look for the next message to send to the sequence, and the rank increment are locked.

The inbound lock step preferably includes locking the sequence-specific mutex stored in the sequence storage.

Upon receipt of an incoming message, the inbound handler preferably checks the sequence correlation value of the sequence of incoming messages and reads the mutex parameters of the sequence before receiving the incoming message. If the mutex is not locked, the inbound handler receives an incoming message. If the mutex is locked, the incoming message waits for the mutex to be released.

More specifically, the mutex is stored in a sequence register included in the sequence storage.

  Advantageously, there is only one mutex per sequence of inbound and outbound handlers. A storage queue in queue storage ensures that for a given sequence, only one message is propagated to the outbound handler until the outbound handler layer has finished processing the messages in that sequence.

  Preferably, the outbound lock step includes locking the sequence-specific mutex stored in the sequence storage.

  Advantageously, if an outbound handler is available, the outbound handler checks the queue storage for messages that can be processed, retrieves the messages, and processes them. If an outbound handler is available, the outbound handler preferably checks the queue storage 850 for messages that can be processed.

If there is a message, this message is automatically modified and processed for the given sequence.

  In one embodiment, when storing incoming messages in sequence storage, specifically overflow storage, the inbound handler sends an acknowledgment message.

  Typically, an acknowledgment message is sent to the originator of the message.

  Advantageously, messages with a message rank greater than the sequence rank are stored in overflow storage and the message sequence is locked unless the message rank matches the sequence rank (ie, the rank of the next message to be processed). Is done.

  Preferably, messages having a message rank greater than the sequence rank are first stored in the overflow storage and then discarded from the overflow storage after reaching the timeout value assigned to the sequence of messages. Alternatively or additionally, messages having a message rank greater than the sequence rank are first stored in the overflow storage and then discarded from the overflow storage after reaching the timeout value assigned to the message.

  In another embodiment, the invention relates to a non-transitory computer readable medium comprising software program instructions, wherein execution of the software program instructions by at least one data processor is an execution of an operation comprising execution of a method according to the invention. .

In another embodiment, the invention relates to a distributed parallel processing system that performs asynchronous message sequencing, the distributed parallel processing system comprising:
A plurality of inbound handlers including at least one data processor, each configured to independently receive a plurality of incoming messages associated with various sequences;
A plurality of outbound handlers including at least one data processor, each independently processing and forwarding a plurality of incoming messages; and a storage layer, said storage layer comprising at least one memory:
A queue storage for storing incoming messages in a state to be transmitted to a plurality of outbound handlers; and a sequence storage, wherein the sequence storage maintains a status of the sequence of incoming messages and updates a sequence status context (802) ), And overflow storage configured to receive messages from inbound handlers and forward them sequentially to queue storage.
The system also determines whether the incoming message is the next message to be processed in order to maintain the message order in this sequence of messages, and performs the following processing executed by at least one data processor: It is configured as follows.
If the sequence status indicates that none of the outbound handlers is currently processing messages in the sequence and it is determined that the incoming message is a message to be processed next to the sequence, the incoming message is queued. And then forward to an outbound handler that is available for processing.
If the sequence status indicates that at least one of the outbound handlers is currently processing messages for the sequence, or if the queue storage already contains messages to be processed for the sequence, or an incoming message Is determined not to be the next message to be processed in the sequence, the incoming message is stored in shared overflow storage and retained for further processing.

  In additional embodiments, storage layer queue storage and sequence storage are implemented in in-memory data or file-based storage. Alternatively, storage layer queue storage and sequence storage are implemented in a client / server storage database.

  Preferably, the checking of the sequence status includes obtaining the status of the sequence based on the sequence correlation value of the sequence.

In another embodiment, the invention is implemented by a computer to process asynchronous messages between at least one server application and at least one client application in a parallel environment having multiple parallel inbound handlers and multiple parallel outbound handlers. With regard to the travel monitoring method to be performed, this method performs the following steps with at least one data processor:
Receiving an incoming message with a sequence correlation value specifying a sequence including the incoming message in one inbound handler of a plurality of parallel inbound handlers;
Check the sequence status of the sequence in the sequence storage;
Determining whether the incoming message is the next message to be processed in order to maintain the message order in the sequence;
Sequence status indicates that none of the multiple outbound outbound handlers is currently processing messages in the sequence, and the incoming message is determined to be the next message to be processed after the sequence If so, send the incoming message to queue storage and then forward it to an outbound handler available for processing;
If the sequence status indicates that at least one of the outbound handlers is currently processing a message of the sequence, or if the queue storage already contains a message to be processed of the sequence, or an incoming message Is determined not to be the next message to be processed in the sequence, the incoming message is stored in the shared overflow storage and retained for further processing;
The message includes data relating to passengers and data including sequence correlation values relating to traffic service references.

  The method of the present invention further includes any one of the following additional functions and steps.

Once processed, the message is forwarded from the outbound handler to at least one of the following: Travel reservation system, airline inventory system, airline e-ticket system, airport departure control system, airport operation system, airline operation system, ground service operation system.

  In one embodiment, the traffic service reference includes at least one of a flight number, date and booking class.

The message in one embodiment specifies at least one of a passenger, a canceled passenger, or an additional passenger.

In one embodiment, each incoming message is given a sequence timeout value to delete incoming messages stored in overflow storage after the sequence timeout value is reached, and the sequence timeout value is the flight departure time or It is activated by either one of the end of flight presentation or the end of promotion.

  In another embodiment, the present invention relates to a non-transitory computer readable medium comprising software program instructions, wherein execution of said software program instructions by at least one data processor comprises execution of operations including execution of said method according to the present invention. Become.

In yet another embodiment, the present invention provides a computer for sequencing delivered asynchronous messages in a distributed parallel system having a plurality of inbound handlers and a plurality of outbound handlers and including at least one processor for processing messages. The method includes the following steps using at least one data processor:
In an inbound handler, receive an incoming message with a sequence correlation value that identifies a sequence including the incoming message, and determine a message rank indicating the order of the incoming message in the sequence;
Check the sequence status of the sequence in the sequence storage:
The sequence status indicates that none of the outbound handlers is currently processing the message of the sequence;
The received incoming message does not have any index indicating the message rank in the sequence and the sequence storage does not yet contain the message to be processed in the sequence; or
Received incoming messages are given an index indicating the message rank in the sequence (equivalent to the sequence rank shown in the sequence storage), the message rank defining the rank of the message in the sequence to be processed next;
It then forwards the incoming message to queue storage, which is then forwarded to an outbound handler that can process it.
If the sequence status indicates that at least one of the outbound handlers is currently processing messages for the sequence, or if the queue storage already contains messages to be processed for the sequence, or received If an incoming message is indexed to indicate the message rank in the sequence (greater than the sequence rank shown in the sequence storage), and the message rank defines the rank of the message in the sequence to be processed next, then Incoming messages are stored in overflow storage for further processing.

The foregoing embodiments of the invention and embodiments of other aspects will be further understood by reference to the following figures.

FIG. 1A is a flowchart illustrating the order of processing synchronous calls or messages between a calling system and a remote system. FIG. 6 is a flow chart illustrating the order of processing asynchronous calls or messages between a calling system and a remote system. 6 is a flowchart illustrating parallel processing of calls or messages in a distributed processing system. 6 is a flowchart illustrating the parallel processing of asynchronous calls or messages in a distributed processing system and the risk of message processing not being performed in order and the risk of data corruption. FIG. 4 is a block diagram illustrating high level sequence management in centralized and shared sequence contexts according to the present invention. 6 is a flowchart illustrating a process for identifying a sequence in a transmission and processing channel according to the present invention. 1 illustrates an asynchronous distributed processing system according to the present invention. FIG. FIG. 4 illustrates the steps of the sequencing process according to the invention, where the inbound handler receives the first message of sequence A. FIG. 6 illustrates another step of the sequencing process according to the invention, where the inbound handler receives the second message of sequence A. FIG. 6 illustrates another step of the sequencing process according to the present invention in which the outbound handler processes the first message of sequence A. FIG. 6 illustrates another step of the sequencing process according to the present invention, in which the outbound handler processes the first message of sequence A. FIG. 4 illustrates the steps of the sequencing process according to the invention with the sequence rearranged.

  The following description is intended for applications in the travel industry, but the present invention can be used for data processing for any application, including travel products such as hotel rooms, rental cars, and rail tickets. The invention is not limited to the following examples.

  According to the present invention, the message processing order is indexed by the message emitter or call system in the asynchronous parallel environment to indicate the rank of each message in the sequence before sending the next message in the given sequence. Defined explicitly, or implicitly by delivering messages sequentially and waiting for the acknowledgment of the message to be delivered.

  The present invention aims to ensure that processing performed independently at the same time adheres to the sequencing order for processing a given set of messages, defined as a sequence.

  In that regard, the method, apparatus and system for performing distributed asynchronous message sequencing according to the present invention perform various operations, which will be briefly described below and will be described in detail later with reference to the drawings. explain.

  Each message or service call in a given sequence is tagged with an interface definition and actually refers to the specific sequence to which it belongs.

The rank of a message or service call within a sequence of messages is one of the following:
• Explicitly provided by the emitter / sender of the message or service call via a sufficient interface. For example, a message includes a field with an index that defines the rank of the message in the sequence; or • Implicitly using the order in which messages or service calls in the sequence are received over time Provided to.

  Once the sequence and message rank or service call order is identified, the sequence needs to be managed appropriately.

  FIG. 5 is a block diagram illustrating high level sequence management in a centralized and shared sequence context according to the present invention. This figure details the major functions and major steps of the system.

  The asynchronous distributed processing system includes an inbound handler 510 that receives incoming messages 501 and an outbound handler 530 that is configured to process the messages and deliver them. The system also includes an overflow storage area 540 that may store a message received from an inbound handler if processing of the message 501 needs to be suspended to maintain the sequence order to which the message belongs.

  Although the system includes a plurality of parallel inbound handlers 510 and a plurality of parallel outbound handlers 530, it should be understood that FIG. 5 is shown briefly for clarity.

  Inbound handlers are also called acceptors or acceptor processes. Accordingly, an inbound handler layer including a plurality of inbound handlers may be referred to as an acceptor layer.

  An outbound handler may be referred to as a processor or delivery process. Thus, an outbound handler layer that includes multiple outbound handlers may also be referred to as a distribution layer.

  The inbound handler 510 is configured to perform any one of the following: Receive messages from emitters such as publishers, validate message integrity, perform sequence lock and status validation, store messages in one of two areas (eg queue storage or overflow area), Respond to the emitter.

  According to an advantageous embodiment, the outbound handler 530 consists of two processes. The first process called the delivery process 531 is configured to execute any one of the following processes. Obtain a message from the storage queue; send the message to the recipient via the communication channel; and stop being used by other processes.

  The second process called the affirmative response process 532 is configured to execute any one of the following processes. Receive acknowledgment from recipient; perform sequence management to place next message into corresponding sequence (storage queue, if any); and stop being used by other processes.

  Thus, the delivery layer formed by the outbound handler 532 is asynchronous that accommodates high scalability requirements. In this way, the system is independent of the waiting time of the recipient. More precisely, the outbound handler retrieves and delivers the first sequence of messages, and then sends another message of the second sequence until the outbound handler receives an acknowledgment for delivery of the first sequence of messages. Means that can be acquired and distributed. Thus, the outbound handler can handle messages from many sequences asynchronously, thereby increasing the number of messages that the system can send while always maintaining the exact order of each sequence.

In accordance with the present invention, a central shared sequence context is implemented in which a state machine is used for each sequence. Each time an incoming message 501 is received by the inbound handler 510, the corresponding sequence context status is checked (520). According to one embodiment, if a corresponding sequence context does not exist, it is created dynamically and transparently. Thus, the present invention does not need to define a sequence in advance in the system, and in this respect is totally dynamic. Further, if the message is not indexed to indicate the rank of the message in the sequence, the message rank is assigned to the message according to the arrival order of the messages:
If the sequence status indicates that the outbound handler layer is waiting for the next message in the sequence (ie, the sequence status is “Waiting”), the incoming message 501 is the standard for the outbound handler 530 Is processed (522) as usual by the target operation (messages can be processed asynchronously);
Alternatively, if the sequence message is currently being processed, that is, if the sequence status indicates “Processing” and the sequence status indicates that the incoming message 501 is subject to processing (524) later, the particular sequence Stored in the overflow storage area 540. The overflow storage area 540 is configured / indexed so that the order of incoming messages is not lost. In this way, the incoming message 501 is retained for further processing and is not available for immediate processing (excluded from the sequence).

  The outbound handler layer receives a message of effectively the correct sequence rank that is processed by standard operations.

  Once the message 501 is processed, the outbound handler layer looks for the next message to process in the sequence of the overflow storage area 540. If such a message is found, it is pushed to the outbound handler layer based on standard processing. If no message is found, the sequence status is returned to the setting of “Waiting”.

  The order of each message in the sequence is maintained. The sequence storage defines a sequence rank that indicates the rank of the next message that must be processed in order to maintain the order of the messages in the sequence. The sequence rank is incrementally updated each time message processing is complete. Therefore, the sequence rank can be regarded as a counter.

  Any incoming message that does not match the sequence rank (ie, the rank of the next message to be processed) is stored in the overflow storage area 540 until the correct message to be processed is received by the inbound handler 510. This means that insert / delete operations are performed in the overflow storage area 540, taking into account the rank of the sequence and the rank of each message.

  Since the message is stored in the overflow storage area 540 and processing is suspended in the sequence, the sequence is not unlocked by the next message in the sequence. Although this situation does not occur frequently, the present invention provides an indicator in the context of a sequence to take action on sequences that are considered expired, such as discarding expired messages and sequences. Provide a dynamic way to leave.

  Some MOMs (Message Oriented Middleware) provide sequencing capabilities by avoiding the use of parallel processing (ie, they only run for a single dequeued consumer). They therefore guarantee the sequence at the expense of scalability. Oracle (R) Advanced Queueing is a typical example.

  Some other MOMs (such as MQ-Series) provide correlator-based sequencing functions, which require that the messages in the sequence to be processed are logically grouped. In addition, there is a limit on the size of the group, and the MOM may have a new limit on the dequeue process.

The distributed parallel processing according to the present invention provides strict sequencing that maintains parallelism and scalability, and has no special restrictions on how messages or service calls are associated or processed by dequeue processing. With the high scalability and fault tolerance of the method, apparatus and system of the present invention, the following is achieved:
• Using the “post and quit” principle, another process (acknowledgment process) is responsible for receiving the acknowledgment, while the message is posted and processed without waiting for the acknowledgment. Perform asynchronous delivery processing, which allows processing with very high message throughput;
And, by performing full distributed processing and removing the similarity between sequences and inbound / outbound handlers, inbound / outbound handlers can handle any sequence of messages.

  The usual method of message sequencing may return to the monoprocess structure, which is critical in fault tolerance and scalability, and has unacceptable limitations. In contrast, the present invention provides many advantages in two stages, inbound handler level and outbound handler level, in distributed parallel processing while keeping the sequence density high. In other words, if the system needs to handle a large number of sequences in parallel, the present invention provides many advantages for parallelization of sequence processing.

There are no prerequisites for storage space and dequeuing for sequence maintenance:
• Inbound handler message acceptance queuing and outbound handler layer message processing queuing are accomplished by the present invention and need not support sequence preservation;
• Parallel storage and retrieval of messages (ie queuing / dequeuing) is fully maintained;
According to a non-limiting embodiment, the queue storage itself can be local to the node, and the overflow storage area remains global (ie shared by all outbound handlers). Since any node can process a given sequence, overflow storage needs to be shared, and overflow storage accesses a unique overflow area and actually queues and dequeues in the storage. If queue storage is not shared by all outbound handlers, it can be dedicated to a single outbound handler or multiple outbound handlers. If queue storage is not shared by all outbound handlers, each message is received by only one local queue storage;
-Since the status of the message is changed and placed in the overflow area, there is no need for an exception queue, and storage of rejection messages is easy.

  According to the method, apparatus, and system of the present invention, a message sequence is processed in a distributed parallel mode by identifying the sequence and identifying the message rank within the sequence. In addition to specifying, sequences are managed and rearranged, including sequence locks and timeouts. These aspects are described in detail below.

In a flow of messages or events shared in a given transmission channel of a sequence, a set of related messages needs to be clearly defined because the sequence takes precedence. FIG. 6 is a flowchart illustrating the process of identifying a sequence in the transmission and processing channel 620 between the Emit system 610 and the process system 630.

  A dedicated parameter is provided for each message primitive included in a given transmission. This dedicated parameter is a sequence identifier and is also called a sequence correlation value. This is typically an alphanumeric value set by the message's emmit system. This parameter is used by each participating component and identifies messages belonging to the same sequence. For example, messages 1-4 are analyzed in transmission channel 620 and identified as messages # 1- # 4. These associated and ordered messages share the same transmission channel 620 and do not continuously track each other. These combine in the transmission channel 620 with messages belonging to other sequences.

  Sequence correlation parameters are defined to ensure that they are not shared by individual conflicting processes in a given transmission and processing chain. In this situation, it must be strictly unique. The definition of the sequence correlation parameter is the responsibility of the business process using this system.

Identification of the message rank in the sequence Messages that need to be processed in a specific order are classified into two types:
For the first type of message, the order or rank in the sequence is known at the time of message creation, and the total number of messages is preferably also known at the time of message creation. A rank number can be assigned. This message rank number is then carried and stored by each process as part of the overall chain until the final process is performed;
The order in the sequence of the second type message is determined at the time of generation. The processing of these second types of messages is usually incremental, that is, each new message (or event) in the process changes the processing result of the previous message in the sequence. In that regard, the message emit system does not know the message rank or number of messages in a given sequence, nor does it know the total number of messages in the sequence.
For simplicity, this description refers to the message rank number of messages in a given sequence as the message rank.

Core Sequence Management As illustrated in FIG. 7, the asynchronous distributed processing system includes:
A plurality of inbound handlers 710, 720,..., 740 form an inbound handler layer. The inbound handler receives the incoming messages 711, 721, 731 and 741, stores them individually in the queue storage 750, and acknowledges that these incoming messages 711, 721, 731 and 741 have been received successfully. , Send to Emit system;
A plurality of outbound handlers 760, 770, ..., 790 form an outbound handler layer. Each outbound handler is configured to retrieve messages from queue storage 750 and process them. The outbound handler is also responsible for forwarding processed messages to the application.

  Inbound handlers are also called acceptors or acceptor processes. An outbound handler may be referred to as a processor or delivery process.

FIG. 8A is an improvement over the embodiment of FIG. 7 illustrating the steps for the inbound handler to sequence the first message of sequence A received from the emit system. With this improvement, an additional component called sequence storage 800 is added as part of the storage layer between multiple inbound handlers 810, 820,..., 840 and multiple outbound handlers 860, 870,. Has been implemented. The sequence storage 800 includes:
A mutex 804, also called a mutex, which is a centralized or shared sequence, ensures that only one handler handles a given sequence (or one with the same sequence correlation value) at a time. If there are conflicts, the first-come-first-served basis;
A centralized or shared sequence status context 802, also called status, maintains the sharing of sequence status between all processes, and the sequence is uniquely identified by the sequence correlation value of the sequence itself. The status can also determine the action to apply for each event in a given sequence;
A “waiting” status means that the next incoming message is queued for delivery;
“Delivering” status means that the next incoming message is held;
Centralized or shared overflow storage 806, also called overflow, or overflow storage area, ensures that only the next message processed by multiple outbound handlers can access it, and others are "pending ( pending) ". Overflow storage is an indexed, ordered sequence message that is not ready to be delivered with respect to the current sequence status.

The three components of status 802, mutex 804, and overflow storage 806 are context processing information and have the same properties as queue storage 850. These are implemented as follows:
• Memory data or file-based storage if all distributed processing is performed on the same node; or • Client / server if distributed processing is performed on several nodes where the server system is a remote system Storage database.

  The greatest consistency between the storage layer and standard message storage is to implement both in a shared RDBMS engine that shares its own transactions.

According to the method, apparatus and system of the present invention:
The queue storage 850 enables the exchange of messages between multiple inbound handlers and multiple outbound handlers and operates independently of the sequence storage 800;
The overflow storage 806 of the sequence storage 800 ensures sequencing of message exchange between the plurality of inbound handlers 810, 820,..., 840 and the plurality of outbound handlers 860, 870,.

FIG. 8A illustrates the incoming message sequencing process of the present invention:
Message 801-1 belongs to sequence correlation value “A” and is received by inbound handler 810;
The inbound handler 810 locks the mutex 804 of sequence “A”, as shown at 812, preventing all inbound or outbound handlers from handling messages with a sequence correlation value of “A”;
The inbound handler 810 checks whether the sequence does not exist in the central status context 802 of the sequence “A”, as shown at 814, whether the sequence is in a “Waiting” status;
The inbound handler 810 sets the status context 802 of the sequence “A” to “Processing” as indicated by 814. The present invention assigns a message rank to the incoming message that is the same as the rank incremented by adding 1 to the previously received message. Since the incoming message is the first message in this sequence, the message rank assigned to the message is set to 1. The message rank is preferably stored in the sequence storage 800, more precisely in the sequence context 802. Inbound handler 810 preferably stores message 851 in queue storage 850, as shown at 816; and • Inbound handler 810 acknowledges the message emitter, releases mutex 804 of sequence “A”, etc. Prepare to receive all incoming messages.

FIG. 8B illustrates the next step in the sequencing process where other incoming messages are received by the system:
The second message 801-2 belongs to the sequence “A” and is received by the inbound handler 820;
The inbound handler 820 locks the mutex 804 of sequence “A”, as shown at 822, and the inbound handler 810, 830, or 840, or the outbound handler handles another message in the sequence “A” prevent;
The inbound handler 820 checks the central status context 802 of the sequence “A” whose sequence status is “Processing”, as indicated at 824. Since the message is not available to any outbound handler, inbound handler 820 stores message 807 in overflow storage 806 as shown at 826.
A message rank value incremented by adding 1 to the previous incoming message is assigned to the incoming message as the message rank. Since the message rank of the previous message is 1, the message rank assigned to the incoming message is 2. In addition, the present invention increments a sequence rank that defines the rank of the next message to be processed in that sequence. If multiple messages from the same sequence are stored in the sequence storage 800, these message ranks allow the system to identify the correct message to be transferred to the queue storage 850. A correct message is a message having a message rank corresponding to the sequence rank defined in the sequence. Advantageously, this applies when the incoming message has a message rank assigned by the emitter, and the incoming message rank is assigned according to the arrival order in the inbound handler;
Inbound handler 820 acknowledges the message emitter, releases the mutex of sequence “A”, and prepares to receive any other incoming message.

FIG. 8C illustrates the next step in the sequencing process, where a message stored in the queue storage queue is dequeued for processing by one of the outbound handlers:
One of the outbound handlers 870 obtains a message 851 of sequence “A” from the queue storage 850 as indicated at 871. According to the invention, this message is automatically the next message in the sequence to be processed. Therefore, the rank is a value obtained by adding 1 to the rank of the last message processed. In this illustrated embodiment, the message stored in the queue storage is the first message in sequence A, so its rank is necessarily “1”;
Outbound handler 870 delivers the message with message rank “1”, as shown at 873, by the relevant recipient or another routing means before further delivery to that recipient. Once the message is sent by step 873, the outbound handler 870 is available for further processing. This continues until an acknowledgment is received from the recipient. For example, the outbound handler 870 can obtain and send another message for another sequence, thereby achieving asynchronous delivery and improving throughput. This can also be acknowledged by the emitter's message. Thus, the response time of the message emitter in step 873 has no limit on the number of messages and processes processed by the outbound handler;
The receiver receives the message from the delivery process 8701 of the outbound handler 870 transmitted in Step 873. In response, the recipient sends an acknowledgment message to the system. An acknowledgment process 8702 of the same outbound handler 870 or an acknowledgment process 8602 of another outbound handler 860 receives the acknowledgment message. This corresponds to step 874 shown in FIG. 8C.

FIG. 8D shows the next step in the sequencing process where the next message is stored in the overflow storage area 806 and forwarded to the queue storage 850 before being forwarded to one of the outbound handlers for processing. The steps shown in FIG. 8D are triggered by receipt of an acknowledgment message 875 in the inbound handler of the system:
Outbound handler 860 checks the sequence rank of status context 802 shown at 862 to determine the rank of the next message to process within the sequence of messages that have been acknowledged. Since the sequence rank is set to “2”, the outbound handler 860 acquires the message “807” of the sequence “A” from the overflow storage area 806 having the message rank “2” indicated by 809. This message 807 is then stored in the queue storage 850 and is available to all outbound handlers. The status context 802 remains “Processing”. The sequence rank is incremented and set to “3”, indicating that the next message to be processed has a message rank equivalent to “3”;
Outbound handler 860 stops and prepares to process another message stored in queue storage 850;
Processing continues until the incoming message has been processed and delivered through all sequences.

  The processes shown in FIGS. 8A to 8D are unified with respect to the presence / absence of the sequence index (indicating the rank of the message in the sequence) of the received incoming message. If it is provided, it is used as a sequencing rank unless ranking is generated by an inbound handler based on the order of reception within the same sequence.

Sequence Reordering In addition to the process shown in detail in FIGS. 8A-8D, where incoming messages are received by the inbound handler according to a strict sequence order, the same process was performed to process messages received out of sequence. The only practical prerequisite is that the message emitter provides an index indicating the rank of the message in the sequence for each message in the same sequence.

  As indicated above, the present invention increments a sequence rank that defines the rank of the next message to be processed in that sequence. When the queue storage 850 can receive a message from a given sequence, the sequence rank is checked. Only messages whose message rank is equivalent to the sequence rank are transferred to the queue storage 850. If a message having a message rank equivalent to the sequence rank is not in the overflow area 806 of the sequence storage 800, processing of this sequence is suspended until a message of the correct rank is received from the inbound handler. Thus, the sequence rank operates like a counter indicating the message to be processed. The sequence rank is preferably stored in the sequence storage 800.

FIG. 8E illustrates the steps of the sequencing process for incoming message 801 with the sequence rearranged:
In addition to the above-described processing performed by the plurality of inbound handlers 810,..., 840, steps related to the rearrangement performed by the index instruction may be added;
The message rank indicated by the message index is compared with the rank of the next message to be processed in order to maintain the sequence order. This rank of the next message to process to maintain the sequence order is preferably indicated in the sequence storage by a sequence rank that is incremented and updated;
If the message rank matches the sequence rank as shown at 818, the incoming message 815 is stored in the queue storage 850 and is available to multiple outbound handlers. The sequence status is set to “Processing”. The sequence rank to be processed is incremented and the processing described above is applied;
If the rank of the message indicated by the index exceeds the sequence rank 819, the message 813 is stored in the overflow storage 806. The sequence status 802 is set to “Pending”. The message is effectively not processed. Sequencing is resumed when a message to be processed with the desired message rank is received by the inbound handler. In that case, the inbound handler stores the corresponding message in the queue storage and sets the sequence status to “Processing”.

  8A-8D, once the outbound handler has finished operating on the message, the outbound handler looks in the overflow storage 806 for messages whose message rank matches the sequence rank (the sequence rank determines the sequence order). To indicate the rank of the next message to be processed). If the corresponding message is found, the message is stored in the queue storage 850. If the corresponding message is not found, the sequence status is “Pending” (the message in this sequence exists in the overflow area, but the message The rank is set to “Waiting” (if there is no message for that sequence in the overflow storage area).

Sequence Lock and Timeout Management As described above, received messages of a given sequence are stored in overflow 806 unless their message rank matches the message to be processed. This is locked for the entire sequence unless the next expected message to be processed is received by the inbound handler.

  In certain embodiments, the present invention ensures that this lock state is limited in time if requested by a process. This process further defines a global sequence timeout value expressed in length (seconds, minutes, days, etc.).

  In other embodiments, the sequence context 802 may include an absolute time value defined as a sequence timeout. Each time an inbound or outbound handler accesses a given sequence context record, that is, if the message belonging to the sequence is being processed in some way and indicates action on the sequence, this absolute time value is the current system time. And the value that is the sum of the sequence timeout periods.

  In yet another embodiment, a timeout sequence collector may be implemented that periodically activates and scans the full list of sequence contexts. With this special implementation, any sequence whose sequence period has expired is detected. This process takes advantage of the sequence timeout value to achieve the selection.

By implementing this, the method, apparatus and system according to the present invention can:
Delete corresponding messages in overflow storage 806 and sequence context 802;
Perform sufficient processing and recording for specific sequence deadline events;
Take items in order or ignore items while waiting for the next until they are found;
・ Sound an alarm.

The present invention has many applications for data processing. However, it is particularly suitable for the following:
A messaging server, such as an AMS (Amadeus Messaging Server), where an application operating as an infrastructure hub of a company, more specifically a software company, handles long-time messaging. In industries that require a booking service, AMS is used in both booking and departure management systems. Sequencing is performed for all teletype traffic. Teletype messages are usually called TTY. TTY Type B is an airline industry standard that exchanges messages over asynchronous channels that process in a strict order in a given functional context. For example, the first message includes a list of passengers, the second list includes suppressed passengers, and the third list includes a list of additional passengers. These passenger lists need to follow the order strictly;
Another area of application is, for example, OHF (OTF high-level framework), which is a middleware software component used in many applications to perform guaranteed asynchronous delivery. The primary use of OHF sequencing is to synchronize between CDB (coupled database) and electronic ticket applications. Often, many changes are made to a coupon within a limited time. These changes need to be transferred to the electronic ticket application or coupon database in the correct order to remain synchronized.

  While the above has primarily focused on travel solutions provided by airlines or air carriers, those skilled in the art will recognize that embodiments of the present invention are not limited to, but not limited to, use of airlines, ships, railways It should be understood that other types of travel styles and travel providers can also be applied, including travel products such as cars, buses and hotels.

  The above illustrates various methods, apparatus, and software from the whole using typical and non-limiting examples, and operates as an embodiment of the present invention. However, various modifications and adaptations will become apparent to those skilled in the relevant art when taken in conjunction with the accompanying drawings and claims. As part of the example, other similar or equivalent processes or algorithms and data representations may be tried by those skilled in the art. Furthermore, the names of each component, function, and algorithm (eg, etc.) are merely descriptive and are not intended to have a limited meaning; these components, functions, and algorithms are You may call it by any name as appropriate. All such similar modifications of the teachings of the present invention are within the scope of embodiments of the present invention.

  Furthermore, some features of embodiments of the present invention are advantageously utilized without the corresponding use of other features. Therefore, the above description is merely illustrative of the principles, teachings, and embodiments of the present invention, without being limited thereto.

  Embodiments of the various techniques described herein may be implemented in digital electronic circuitry, computer hardware, handheld electronic hardware, firmware, software, or combinations thereof. This embodiment is executed by a data processing device (eg, by a programmable processor, computer, tablet, or multiple computers) or as a program or software product to control the operation of the data processing device, ie It may be implemented as a computer program tangibly embodied in an information carrier (eg in a machine-readable storage device or in a propagated signal). Programs such as the computer programs described above can be written in any form of programming language, including compiled or interpreted languages, stand-alone programs, modules, components, subroutines, and other computing environments Can be deployed in any form including devices suitable for use. The program is distributed on one computer or tablet, multiple computers or tablets, one site, or multiple sites, interconnected via a communication network or wireless network, and deployed and executed Is possible.

  Examples of suitable processors for executing computer programs include general and special purpose microprocessors and one or more processors of various types of digital computers, tablets, or electronic devices. Generally, a processor receives instructions and data from a read-only memory (ROM), a random access memory (RAM), or both. The computer components comprise at least one processor for executing instructions and one or more memory devices for storing instructions and data. In general, a computer or electronic device has one or more mass storage devices (eg, magnetic disk, magneto-optical disk, optical disk) or is effectively coupled to receive, transmit, and transmit and receive data. Has been.

  The present embodiment includes a computing system that includes a back-end component (eg, a data server), a computing system that includes a middleware component (eg, an application server), a computing system that includes a front-end component (eg, a graphical user interface, or A client computer that has a web browser through which a user can interact with an implementation program) or implemented in a computing system that combines such backend, middleware, or frontend components Also good. The components may be interconnected by any form or medium of digital data communication, such as a communication network, a wireless network, a telecommunications network, etc. Examples of communication networks or telecommunications networks include a local area network (LAN) and a wide area network (WAN) (eg, a wireless network such as the Internet or a WiFi network).

  Although specific features implemented have been described herein, those skilled in the art will appreciate that many modifications, alternatives, changes, and equivalents will occur. Accordingly, it is to be understood that the claims herein contain all such modifications and changes in accordance with the spirit and scope of the embodiments of the present invention.

Claims (21)

Distributed asynchronous message sequencing by a computer using at least one data processor in a distributed parallel system having a plurality of inbound handlers forming an inbound handler layer and a plurality of outbound handlers forming an outbound handler layer How to:
The inbound handler of any of the plurality of inbound handlers receiving an incoming message that includes a sequence correlation value that identifies a sequence that includes the incoming message;
Checking the sequence status of the sequence in shared sequence storage; and determining whether the incoming message is the next message to process in order to maintain the order of messages in the sequence;
If the sequence status indicates that none of the outbound handler layer's outbound handlers are currently processing messages in the sequence, and it is determined that the incoming message is a message to be processed next to the sequence, the incoming Forward the message to queue storage, then get the message by an outbound handler available in the outbound handler layer for processing,
If the sequence status indicates that at least one of the outbound handlers of the outbound handler is currently processing messages of the sequence, or if the queue storage already contains messages to be processed of the sequence, Alternatively, if it is determined that the incoming message is not a message to be processed next in the sequence, the incoming message is stored in the memory of the shared overflow storage and retained for further processing. The method of claim 1, comprising determining whether an incoming message is a next message to be processed in order to maintain the order of the messages in the sequence:
Determining a message rank indicating the order of incoming messages in the sequence; and comparing the message rank with a sequence rank defining a rank of the next message to be processed in the sequence;
If the message rank is the same as the sequence rank, the message is determined to be the next message to be processed in order to maintain the order of the messages in the sequence;
If the message rank is not the same as the sequence rank, the message is not determined to be the next message to be processed in order to maintain the order of the messages in the sequence.   3. The method according to claim 1 or 2, wherein when a message of a given sequence is completed in an outbound handler, the sequence rank of the given sequence is incremented and the sequence rank of the sequence is incremented. When the overflow storage includes a message having a message rank that is the same as the incremented sequence rank, the method forwards the message to the queue storage.   4. The method according to claim 2, wherein the step of determining the message rank when the received incoming message is not given an index indicating the message rank of the sequence indicates the rank of the incoming message in the sequence. Assigning a message rank to the incoming message and storing the assigned message rank in sequence storage.   5. The method of claim 4, wherein the assigned message rank corresponds to the rank of the last message received at any one place in the inbound handler, incremented by adding 1 to the sequence. Method.   The method according to any one of claims 1 to 3, wherein an index indicating a message rank in a sequence is assigned to an incoming message received by an inbound handler.   The method according to any one of claims 2 to 6, wherein the queue storage does not contain any message of the sequence of incoming messages, and the message rank of the incoming message is indicated in the sequence storage. If greater than the rank, the incoming message is stored in overflow storage until the sequence rank is incremented and is equal to the message rank of the incoming message.   The method according to any one of claims 1 to 7, wherein the outbound handler operates asynchronously, so that a message can be transmitted, and after transmitting the message, an acknowledgment is received from the recipient of the message. Until another process is possible.   The method according to any one of claims 1 to 8, wherein the outbound handler includes a delivery process for transmitting a message to the receiver and an acknowledgment process for receiving an acknowledgment from the receiver, and the delivery process and the positive confirmation. The way the behavior is independent.   10. A method as claimed in any preceding claim, wherein all inbound handlers are prevented from receiving another message in the sequence from receiving an incoming message until checking. Performing the step of locking the inbound until the incoming message is stored in the sequence storage or being sent to the queue storage, and the step of locking the inbound is dedicated to the sequence stored in the sequence storage. A method, comprising locking the mutex.   11. The method according to any one of claims 1-10, wherein after an incoming message is transferred from queue storage to an outbound handler, all other outbound handlers receive another message in the sequence. Performing the locking outbound step, which is prevented until message processing is complete, and the locking outbound step includes locking the sequence-specific mutex stored in sequence storage. .   12. The method according to claim 11, wherein when an outbound handler is available, the mutex of the sequence of messages in queue storage is checked and the message is obtained only if the mutex is not locked. ,Method.   13. A method as claimed in any one of claims 2 to 12, wherein a message having a message rank greater than the sequence rank is first stored in the overflow storage, and then the message sequence and / or the message A method that is discarded from overflow storage after reaching the assigned timeout value.   A non-transitory computer readable medium containing software program instructions, wherein execution of the software program instructions by at least one data processor comprises execution of the method of any one of claims 1-13. Computer readable medium. A distributed parallel processing system for sequencing asynchronous messages:
An inbound handler layer comprising a plurality of inbound handlers including at least one data processor, each of the plurality of inbound handlers configured to independently receive a plurality of incoming messages associated with various sequences An inbound handler layer comprising a plurality of inbound handlers; and
An outbound handler layer comprising a plurality of outbound handlers including at least one data processor, each of the plurality of outbound handlers configured to independently process and forward a plurality of incoming messages; A distributed parallel processing system that includes an outbound handler layer, including an outbound handler; the system:
Including a storage layer including at least one memory;
-Queue storage to store incoming messages that are ready to be transmitted to multiple outbound handlers; and
-Includes shared sequence storage; the shared sequence storage is:
-A sequence status context to maintain and update the status of each sequence of incoming messages; and
-Shared overflow storage configured to receive messages from inbound handlers and forward them sequentially to queue storage;
The system also determines whether the incoming message is the next message to be processed in order to maintain the message order in this sequence of messages and performs the following steps executed by at least one data processor: A system configured as follows:
-If the sequence status indicates that none of the outbound handler layer's outbound handlers are currently processing messages in the sequence and it is determined that the incoming message is a message to be processed next to the sequence, Sending the incoming message to queue storage and then retrieving the message by an outbound handler available for processing; and
-If the sequence status indicates that at least one of the outbound handler layer's outbound handlers is currently processing messages of the sequence, or if the queue storage already contains messages to be processed of the sequence Or, if it is determined that the incoming message is not the next message to be processed in the sequence, the incoming message is stored in an overflow storage memory and held for further processing. A distributed parallel processing system for sequencing asynchronous messages. Using at least one data processor to process asynchronous messages between at least one server application and at least one client application in a parallel environment having a plurality of parallel inbound handlers and a plurality of parallel outbound handlers; A computer-implemented travel monitoring method comprising:
Receiving an incoming message in one of the plurality of inbound handlers with a sequence correlation value that identifies a sequence that includes the received message;
Checking the sequence status of the sequence in sequence storage;
Determining whether the incoming message is the next message to be processed in order to maintain the order of the messages in the sequence;
When the sequence status indicates that none of the plurality of outbound handlers is currently processing the message of the sequence, and it is determined that the incoming message is a message to be processed next to the sequence , Send the incoming message to queue storage, then forward it to an outbound handler available for processing,
If the sequence status indicates that at least one of the outbound handlers is currently processing a message of the sequence, or if the queue storage already contains a message to be processed of the sequence, or an incoming message Is not a message to be processed next in the sequence, the incoming message is stored in shared overflow storage and held for further processing;
The travel monitoring method, wherein the message includes data relating to passengers and data including sequence correlation values relating to traffic service references.   17. The method of claim 16, wherein the processed message is sent from an outbound handler to a travel reservation system, an airline inventory system, an airline e-ticket system, an airport departure management system, an airport operation system, and an airline. And a method for transferring to at least one of a ground service operation system.   18. A method according to any one of claims 16 and 17, wherein the traffic service reference includes at least one of a flight number, a date and a booking class.   19. A method according to any one of claims 16 to 18, wherein the message describes any one of a passenger, a canceled passenger, or an additional passenger.   20. The method according to any one of claims 16 to 19, wherein a sequence timeout value triggered by any one of a flight departure time, a flight presentation end, or a promotion end is reached. A method in which each incoming message is given a sequence timeout value to later delete incoming messages stored in overflow storage.   21. A non-transitory computer readable medium comprising software program instructions, wherein execution of software program instructions by at least one data processor comprises execution of operations comprising the method of any one of claims 16-20. Non-transitory computer readable medium.

Images